Many computing applications such as computer games, multimedia applications, or the like use controls to allow users to manipulate game characters or other aspects of an application. Conventionally, such controls are input using, for example, controllers, remotes, keyboards, mice, or the like. Unfortunately, such controls can be difficult to learn, thus creating a barrier between a user and such games and applications. Furthermore, such controls may be different than actual game actions or other application actions for which the controls are used. For example, a game control that causes a game character to swing a baseball bat may not correspond to an actual motion of swinging the baseball bat. Recently, cameras have been used to allow users to manipulate game characters or other aspects of an application without the need for conventional handheld game controllers. Further, such cameras have enabled representation of users to be displayed within a game or other application. Such computing systems have often relied on skeletal tracking (ST) techniques to detect motion or other user behaviors as well as to produce representations of users for display. However, while useful for detecting and displaying representations of users in certain situations, ST techniques have proven to be unreliable for detecting and displaying representations of users in other situations. For example, ST techniques are typically unreliable for detecting and displaying representations of a user where the user is laying or sitting on or near the floor.